Measure Values, Toil, and Trouble: Avoiding the ‘Dead on Delivery’ Dashboard

Tableau has this nifty feature: measure values.  Its swell companion, measure names, makes doing certain types of analyses painless.  This feature is a bit of a double-edged sword, in that it makes us comfortable, sometimes even – dare I say it – numb to badly shaped data.

Here’s what I mean:

I got measure values, I got rhythm!

We can see it’s some type of ratings data.  If I sit and decipher my very long measure names, I’ll figure out I have some demographic information (age ranges) and types of ratings (15 minute versus Full Show) all clustered in with my population.  I can look at this a few different ways:

Holy numbers batman!

As I look through the data manually, I can see a million numbers and not a whole lot of words.  We like numbers, right?  Numbers make sense.

Except, JUST numbers don’t help us make sense of our data.  I can make a few charts.

I can spend an absurd amount of time calculating across columns.

And, of course, I can toss this all on a dashboard…

LOOK MA!  A dashboard! And in the past, I would’ve called it done.  Because historically, with legacy BI, this is all I could do.  In the past, the wall separated me from my data, making work beyond this difficult and not worth the value.

Yet, I sure don’t seem too happy about this dashboard.  For one, I want to be able to break out full-show versus partial-show data, and I want to be able to compare across age groups.  I could spend (even more) time creating more calculations and parameters to do this, but frankly, I want to use Tableau like Tableau, not as a calculations machine.  This dashboard is dead on delivery.  Any further questions ensure I’ll find the nearest escape route.  And you better believe I brought my parachute and hoverboard to this meeting.

Don’t get me wrong, I was able to work with this data.  It wasn’t great and I hacked my way through this.  I could continue to hack my way through this.  But, I’m not.  Tableau is designed to allow me seamless discovery of my data.  I shouldn’t feel like my data is behind a wall.  But this has hardly been seamless for one good reason – Tableau likes dimensions.  And the real data is hidden in measure values and measure names (AKA, the wall of numbers).  That’s a pretty good indicator it’s time to reshape the data.

Let’s look at this.  I’ve pulled out my table and started grouping.

For this exercise, I ended up reformatting it in Excel.  I decided I was really dealing with only 2 numbers: population and rating.  All other numbers are what I’d call derived or non-natural grain.  So, I grouped all the other info as dimensions.  My newly created dimensions, rating type and demographics, would provide me categories for my data, allowing me to slice and dice down to this level of information.  I copied and pasted until I effectively unioned my data into this long narrow format.  Rough and gritty, but it works for this example.

With the data sorted, I pulled this back in to Tableau.

Right away, I can see my column names are shorter – YIPPEE!  I have more fields above the measures line than below it (another win!) and…

I now have a lot more words in this data set.  You’ll see rows repeat and my scrollbar is in the right place (vertical, not horizontal).  Okay, let’s put this through the Tableau Toddler test.  Can we finger paint with this?

Right away, I notice a difference.  First, I don’t feel like I’m locked in on bar charts galore.

Toddler finger painting – BUBBLES!

Now, I’m spending my time exploring the data, analyzing it, and understanding it.  In fact, I ended up only making one calculation for this entire dashboard and it was solely for the purpose of making a tidy legend.  I could’ve gone without it, but some of us have circle obsessions.

And building a dashboard?

Data shape doesn’t affect only your calculations or ease of working – it influences the analysis, which flows down the line to the dashboard.  This dashboard is interactive.  Hovering the 2 primary charts highlights across the others and filters the number of viewers.  I chose to allow rating type to be a filter, but I could also make that a chart.  This is by no means a fancy or complex dashboard, but it lets me see ratings and viewers quickly and without having to hover or get the ruler.  I put it together in far less time than the other and had more fun doing it.

Does this mean you’ll never use measure values or that it’s bad?  Of course not!  I use it often.  But, there’s a tipping point and when you reach it, it inhibits your analysis.

Is it perhaps time to 🎯reshape the data🎯?  Let’s toss some darts at this one!

  1. Where are most of the fields?
    • On top in dimensions: good start.
    • Buried in measures: not promising. 🎯
  2. What types of calculations are you doing across columns?
    • Percentages, ratios, differences, and some creation of sub-categories: excellent!
    • Mostly adding to get totals: Yikes! 🎯
  3. What types of dimensions are you using?
    • Various ones, including dates, several primary fields, occasionally measure names.
    • Measure names by far! Maybe some others. 🎯
  4. What are you spending your time doing?
    • Exploring, creating some calcs, putting together dashboards
    • Writing 900 calculations and grumbling. Isn’t that what you do? 🎯
  5. How do you feel?
    1. Happy, like a Tableau Toddler. Tunes may be from Guardians of the Galaxy.
    2. Angsty, like an Adolescent Analyst. Where’s my Nirvana playlist? 🎯